Event-based evaluation of IMERG for extreme precipitation in China: Insights from an error decomposition approach

Chinese mainland Using hourly precipitation data from over 2100 stations across China, this study applies an event-based error decomposition method to systematically evaluate IMERG’s performance in capturing extreme precipitation processes and investigates the influence of different climate regions, seasons, and topographic conditions on error components. Precipitation occurs in discrete “events” with defined start time, duration, and intensity. Traditional evaluation methods oversimplify the complexity of precipitation dynamics. Therefore, assessing the accuracy of IMERG in capturing extreme precipitation from an event-based perspective is essential for Chinese mainland. Results reveal that IMERG generally overestimates event duration (ED), total precipitation (Esum), and frequency (EF), while underestimating maximum precipitation (Emax). The Total bias is predominantly positive, particularly in humid and semi-humid areas, where False bias dominates. False-Event and Miss-Event significantly contribute to the cumulative extreme precipitation error. IMERG often detects events prematurely, and accuracy is affected by geography and event intensity. Across four climate regions, False-Event is the dominant error source. Seasonal analysis shows False-Event and Miss-Event together contribute 65.3 %–72.4 % of total error. Although contributions vary across elevations, False-Event remains the largest, while Hit Negative dominates individual events. • An event-based error decomposition method was applied for the first time in China. • New method evaluates IMERG's performance in capturing extreme precipitation. • False-Event bias significantly affect errors in extreme precipitation. • IMERG detects extreme event start and end times earlier than actual. • Hit Negative bias causes the largest error in extreme events with elevation.